1
00:00:02,520 --> 00:00:14,320
Speaker 1: Bloomberg Audio Studios, podcasts, radio news.

2
00:00:16,320 --> 00:00:19,680
Speaker 2: Welcome to our Bloomberg radio and television audiences worldwide. We

3
00:00:19,720 --> 00:00:25,239
Speaker 2: go right now to a conversation with Matt Garman, AWS CEO. Matt,

4
00:00:25,400 --> 00:00:28,280
Speaker 2: it's good to catch up. It has been basically one

5
00:00:28,360 --> 00:00:31,200
Speaker 2: year that you've been in the role as AWS CEO.

6
00:00:32,080 --> 00:00:34,840
Speaker 2: Is a place to start what has been the biggest

7
00:00:35,240 --> 00:00:37,960
Speaker 2: achievement in that time for AWS.

8
00:00:39,040 --> 00:00:40,839
Speaker 3: Yeah, thanks for having me on. It's nice to be

9
00:00:40,880 --> 00:00:44,640
Speaker 3: here again. Yeah, it's been a fantastic year of innovation.

10
00:00:45,000 --> 00:00:48,760
Speaker 3: It's really been incredible and as I look out there,

11
00:00:49,120 --> 00:00:50,920
Speaker 3: one of the things that I've been most excited about

12
00:00:51,040 --> 00:00:54,720
Speaker 3: is how fast our customers are innovating and ten adopting

13
00:00:54,840 --> 00:00:57,360
Speaker 3: many of the new technologies that we have. And as

14
00:00:57,400 --> 00:01:00,880
Speaker 3: you think about customers that are on this cloud mication journey,

15
00:01:01,120 --> 00:01:02,840
Speaker 3: many of them have been doing that for over the

16
00:01:02,960 --> 00:01:05,920
Speaker 3: last several years, but this year in particular, that we've

17
00:01:05,959 --> 00:01:09,800
Speaker 3: really seen an explosion of AI technologies, of agentic technologies,

18
00:01:10,160 --> 00:01:14,080
Speaker 3: and increasingly we're seeing more and more customers move their

19
00:01:14,240 --> 00:01:17,240
Speaker 3: entire estates into the cloud and AWS.

20
00:01:17,240 --> 00:01:18,839
Speaker 4: So it's been really fun to see.

21
00:01:18,880 --> 00:01:21,880
Speaker 3: It's been an incredible pace of technology and it's been

22
00:01:21,880 --> 00:01:22,920
Speaker 3: a really fun first year.

23
00:01:23,959 --> 00:01:26,640
Speaker 2: The moment that investors kind of sat up and paid

24
00:01:26,680 --> 00:01:32,000
Speaker 2: attention was when Amazon said that it's AI business was

25
00:01:32,040 --> 00:01:35,040
Speaker 2: at a multi billion dollar run rate in terms of sales.

26
00:01:35,440 --> 00:01:38,120
Speaker 2: What we don't understand as well is what proportion of

27
00:01:38,160 --> 00:01:40,040
Speaker 2: that is AWS infrastructure?

28
00:01:41,120 --> 00:01:43,760
Speaker 3: Yeah, that is AWS, right, And so the key is

29
00:01:43,800 --> 00:01:47,640
Speaker 3: that's a mix of customers running their own models. Some

30
00:01:47,680 --> 00:01:50,480
Speaker 3: of that is on Amazon Bedrock, which is our own

31
00:01:50,520 --> 00:01:53,800
Speaker 3: hosted models, where we have first party models like Amazon Nova,

32
00:01:53,840 --> 00:01:55,720
Speaker 3: as well as many of the third party models like

33
00:01:55,760 --> 00:01:59,160
Speaker 3: Anthropics models, and some of those are applications things like

34
00:02:00,120 --> 00:02:04,840
Speaker 3: on q which helps people do automated software development, as

35
00:02:04,880 --> 00:02:07,040
Speaker 3: well as a host of other capabilities, and so there's

36
00:02:07,080 --> 00:02:08,799
Speaker 3: a mix of that, and I think part of the

37
00:02:08,840 --> 00:02:11,440
Speaker 3: most interesting thing about being at a multi billion dollar

38
00:02:11,480 --> 00:02:14,400
Speaker 3: run rate is we're at the very earliest stages of

39
00:02:14,440 --> 00:02:17,919
Speaker 3: how AI is going to completely transform every single customer

40
00:02:17,960 --> 00:02:18,360
Speaker 3: out there.

41
00:02:18,680 --> 00:02:20,760
Speaker 4: And we talk to customers and we look at where the.

42
00:02:20,760 --> 00:02:24,880
Speaker 3: Technology landscape is, and we firmly believe that every single business,

43
00:02:24,919 --> 00:02:27,840
Speaker 3: every single industry, and really every single job is going

44
00:02:27,919 --> 00:02:30,960
Speaker 3: to be fundamentally transformed by AI. And I think we're

45
00:02:31,000 --> 00:02:33,440
Speaker 3: starting to see the early start the stages of that.

46
00:02:33,639 --> 00:02:35,880
Speaker 3: But again we're just at the very earliest stages that

47
00:02:35,919 --> 00:02:37,800
Speaker 3: I think what's going to be possible, and so that

48
00:02:37,880 --> 00:02:40,200
Speaker 3: multi billion dollar business that we have today is really

49
00:02:40,320 --> 00:02:41,000
Speaker 3: just the start.

50
00:02:42,080 --> 00:02:46,119
Speaker 2: Can you give me a generative AI revenue number.

51
00:02:47,000 --> 00:02:49,239
Speaker 4: For the world or for awls?

52
00:02:49,280 --> 00:02:52,080
Speaker 2: Are you guys for AWS? Maybe Amazon as a whole.

53
00:02:52,240 --> 00:02:55,240
Speaker 3: Yeah, Like I said, we are in multiple billions of dollars,

54
00:02:55,600 --> 00:02:59,240
Speaker 3: and that's for customers using AWS. We also use lots

55
00:02:59,280 --> 00:03:02,160
Speaker 3: of generative AI inside of Amazon for a wide range

56
00:03:02,200 --> 00:03:05,040
Speaker 3: of things. We use it to optimize our fulfillment centers.

57
00:03:05,520 --> 00:03:07,240
Speaker 3: We use it when you go to the retail site

58
00:03:07,280 --> 00:03:11,680
Speaker 3: to summarize reviews, or to help customers find products in

59
00:03:11,680 --> 00:03:15,280
Speaker 3: a faster and more interesting way. We use AI in

60
00:03:15,480 --> 00:03:19,160
Speaker 3: Alexa in our new Alexa Plus offering, where we conversationally

61
00:03:19,200 --> 00:03:23,680
Speaker 3: talk to customers through the Alexa interface and help them

62
00:03:23,680 --> 00:03:26,640
Speaker 3: accomplish things through voice that they were never able to

63
00:03:26,680 --> 00:03:30,280
Speaker 3: do before. So every single aspect of what Amazon does

64
00:03:31,080 --> 00:03:35,280
Speaker 3: leverages AI, and our customers are exactly the same. Customers

65
00:03:35,280 --> 00:03:38,600
Speaker 3: are looking to AWS to completely change, whether it's their

66
00:03:38,600 --> 00:03:42,880
Speaker 3: contact centers through something like Amazon Connect where it shows

67
00:03:43,000 --> 00:03:45,320
Speaker 3: AI capabilities so that you don't have to go program

68
00:03:45,360 --> 00:03:48,000
Speaker 3: it all the way down to our custom chips or

69
00:03:48,080 --> 00:03:51,120
Speaker 3: Nvidia processors or anything where customers at the metal are

70
00:03:51,120 --> 00:03:53,520
Speaker 3: building their own models. We have the whole range of

71
00:03:53,520 --> 00:03:57,000
Speaker 3: people that are building AI on top of AWS as

72
00:03:57,040 --> 00:03:58,400
Speaker 3: well as Amazon themselves.

73
00:03:59,280 --> 00:04:04,520
Speaker 2: We always credit AWS is being number one hyperscala. But

74
00:04:04,960 --> 00:04:07,440
Speaker 2: just what you said there about what the client's using

75
00:04:08,400 --> 00:04:12,480
Speaker 2: the silicon level through to capacity, it would really help

76
00:04:12,560 --> 00:04:16,560
Speaker 2: if you could proportionately tell me what percentage of workloads

77
00:04:16,560 --> 00:04:20,360
Speaker 2: are being run for training and which proportion of workloads

78
00:04:20,480 --> 00:04:21,560
Speaker 2: being run for inference.

79
00:04:21,960 --> 00:04:25,039
Speaker 4: Sure, yeah, and that changes over time. I think.

80
00:04:25,040 --> 00:04:28,599
Speaker 3: Look as we progress over time, more and more of

81
00:04:28,680 --> 00:04:31,240
Speaker 3: the AI workloads are being inference. I'd say in the

82
00:04:31,320 --> 00:04:34,600
Speaker 3: early stages of AI and general of AI, a lot

83
00:04:34,600 --> 00:04:36,880
Speaker 3: of that usage was dominated by training as people were

84
00:04:36,880 --> 00:04:40,120
Speaker 3: building these very large models with small amounts of usage.

85
00:04:40,160 --> 00:04:42,599
Speaker 3: Now the models are getting bigger and bigger, but the

86
00:04:42,720 --> 00:04:45,320
Speaker 3: usage is exploding at a rapid rate, and so I

87
00:04:45,480 --> 00:04:48,880
Speaker 3: expect that over the fullness of time, eighty percent, ninety percent,

88
00:04:48,960 --> 00:04:51,080
Speaker 3: the vast majority of usage is going to be in

89
00:04:51,240 --> 00:04:54,160
Speaker 3: inference out there, and really and just for all those

90
00:04:54,200 --> 00:04:58,159
Speaker 3: out there, inference It really is how AI is embedded

91
00:04:58,240 --> 00:05:01,080
Speaker 3: in the applications that everybody uses. And so as we

92
00:05:01,120 --> 00:05:03,360
Speaker 3: think about our customers building, you know, there's a small

93
00:05:03,440 --> 00:05:05,560
Speaker 3: number of people who are going to be building these models,

94
00:05:05,880 --> 00:05:08,520
Speaker 3: but everyone out there is going to use inference as

95
00:05:08,560 --> 00:05:11,599
Speaker 3: a core building block in everything they do. And every

96
00:05:11,640 --> 00:05:14,280
Speaker 3: application is going to have inference, and already is starting

97
00:05:14,320 --> 00:05:17,520
Speaker 3: to see inference built in to every application. And we

98
00:05:17,600 --> 00:05:20,320
Speaker 3: think about it as just the new building block. It's

99
00:05:20,400 --> 00:05:22,480
Speaker 3: just like compute, it's just like storage, it's just like

100
00:05:22,520 --> 00:05:23,200
Speaker 3: a database.

101
00:05:23,520 --> 00:05:25,039
Speaker 4: Inference is a core building block.

102
00:05:25,080 --> 00:05:26,880
Speaker 3: And so as you talk to people who are building

103
00:05:26,920 --> 00:05:30,039
Speaker 3: new applications, they don't think about it as AI is

104
00:05:30,040 --> 00:05:32,440
Speaker 3: over here and my application is over here. They really

105
00:05:32,440 --> 00:05:35,520
Speaker 3: think about AI is embedded in the experience. And so

106
00:05:35,680 --> 00:05:38,120
Speaker 3: it's increasingly I think it's going to be difficult for

107
00:05:38,200 --> 00:05:40,080
Speaker 3: people to say what part of your revenue is going

108
00:05:40,080 --> 00:05:42,360
Speaker 3: to be driven by AI. It's just part of the

109
00:05:42,400 --> 00:05:44,479
Speaker 3: application that you're building, and it's going to be a

110
00:05:44,520 --> 00:05:46,839
Speaker 3: core part of that experience, and it's going to deliver

111
00:05:47,200 --> 00:05:51,160
Speaker 3: lots of benefits from efficiency, from capabilities, and from user

112
00:05:51,200 --> 00:05:53,480
Speaker 3: experience for all sorts of applications and industries.

113
00:05:54,279 --> 00:05:57,640
Speaker 2: But present day, it's fair to say majority is still training.

114
00:05:58,080 --> 00:06:00,480
Speaker 3: No, I think that at this point more definitely more

115
00:06:00,560 --> 00:06:02,000
Speaker 3: usage as inference than training.

116
00:06:02,960 --> 00:06:06,080
Speaker 2: We want to welcome our radio and television audiences around

117
00:06:06,080 --> 00:06:09,560
Speaker 2: the world. We're speaking to AWS CEO Matt Garman, who

118
00:06:09,600 --> 00:06:14,279
Speaker 2: officially next week celebrates one year in that role leading AWS.

119
00:06:14,680 --> 00:06:19,280
Speaker 2: A new metric that has been discussed, particularly this earning season.

120
00:06:19,320 --> 00:06:22,600
Speaker 2: We discussed it with Nvidia CEO Jensen one this week

121
00:06:22,960 --> 00:06:27,640
Speaker 2: is token growth and tokenization. Has AWS got a metric

122
00:06:27,680 --> 00:06:28,559
Speaker 2: to share on that front?

123
00:06:29,080 --> 00:06:30,760
Speaker 3: I don't have any metrics to share on that front,

124
00:06:30,760 --> 00:06:32,640
Speaker 3: but I think it's one of the measures that we

125
00:06:32,680 --> 00:06:35,120
Speaker 3: can look at as the numbers of tokens that are

126
00:06:35,120 --> 00:06:37,279
Speaker 3: being served out there, but it's not the only one,

127
00:06:37,520 --> 00:06:40,120
Speaker 3: and I increasingly think that people are going to be

128
00:06:40,160 --> 00:06:43,920
Speaker 3: thinking about these things differently. Tokens are a particularly interesting

129
00:06:44,720 --> 00:06:47,080
Speaker 3: thing to look at when you're thinking about text generation,

130
00:06:47,560 --> 00:06:49,640
Speaker 3: but not all things are created equal.

131
00:06:49,680 --> 00:06:51,200
Speaker 4: I think, particularly as you think about.

132
00:06:51,040 --> 00:06:54,800
Speaker 3: AI reasoning models, the input and output tokens don't necessarily

133
00:06:56,240 --> 00:06:59,160
Speaker 3: talk about the work that's being done, and increasingly you're

134
00:06:59,200 --> 00:07:01,799
Speaker 3: seeing models can do work for a really long period

135
00:07:01,839 --> 00:07:05,080
Speaker 3: of time before they output tokens, and so you're having

136
00:07:05,120 --> 00:07:09,240
Speaker 3: these models that can sometimes think for hours at a time. Right,

137
00:07:09,279 --> 00:07:12,000
Speaker 3: you ask these things to go and actually do research

138
00:07:12,000 --> 00:07:13,760
Speaker 3: on your behalf. They can go out to the internet,

139
00:07:13,920 --> 00:07:16,920
Speaker 3: they can pull information back, they can synthesize, they can

140
00:07:17,000 --> 00:07:19,920
Speaker 3: redo things. If you think about coding and que developer,

141
00:07:20,560 --> 00:07:23,160
Speaker 3: we're seeing lots of coding where it goes and actually

142
00:07:23,480 --> 00:07:27,120
Speaker 3: reasons and does iterations and iterations and improves on itself,

143
00:07:27,320 --> 00:07:30,400
Speaker 3: looks at what it's done, and then eventually outputs the

144
00:07:30,520 --> 00:07:32,800
Speaker 3: end result. And so at some point kind of the

145
00:07:33,320 --> 00:07:35,920
Speaker 3: final output token is not really the best measure of

146
00:07:35,960 --> 00:07:38,920
Speaker 3: how much work is being done. If you think about images,

147
00:07:38,960 --> 00:07:41,280
Speaker 3: if you think about videos, there's a lot of content

148
00:07:41,360 --> 00:07:43,920
Speaker 3: that's being created and a lot of thought that's being done.

149
00:07:43,920 --> 00:07:46,000
Speaker 4: And so tokens are one aspect of it.

150
00:07:46,040 --> 00:07:48,240
Speaker 3: And it's an interesting measure, but I don't think it's

151
00:07:48,280 --> 00:07:52,360
Speaker 3: the only measure to look at. Although they are rapidly increasing.

152
00:07:53,840 --> 00:07:58,440
Speaker 2: Project RAY near Massive Custom Server Design project. Yeah, what

153
00:07:58,560 --> 00:08:02,040
Speaker 2: is the operational statu and latest on project right now?

154
00:08:02,200 --> 00:08:05,760
Speaker 3: Yeah, So we're incredibly excited about so project right here

155
00:08:06,880 --> 00:08:09,840
Speaker 3: is a collaboration that we have with our partners at

156
00:08:09,840 --> 00:08:14,400
Speaker 3: Anthropic to build the largest compute cluster that they'll use

157
00:08:14,440 --> 00:08:17,400
Speaker 3: to train their next generation of their claud models, and

158
00:08:18,080 --> 00:08:20,680
Speaker 3: Anthropic has the very best models out there today. Claude

159
00:08:20,680 --> 00:08:23,320
Speaker 3: four just launched, I think it was last week, and

160
00:08:23,800 --> 00:08:28,400
Speaker 3: it's been getting incredible adoption out there from our customer base.

161
00:08:29,720 --> 00:08:31,880
Speaker 3: Anthropic is going to be training their next version of

162
00:08:31,920 --> 00:08:35,280
Speaker 3: their model on top of Trainium two, which is Amazon's

163
00:08:35,280 --> 00:08:40,440
Speaker 3: custom built accelerator processors purpose built for AI workloads, and

164
00:08:40,480 --> 00:08:44,120
Speaker 3: we're building one of the largest clusters ever released. It's

165
00:08:44,120 --> 00:08:47,160
Speaker 3: an enormous cluster, more than five times the size of

166
00:08:47,200 --> 00:08:50,200
Speaker 3: the cluster compared to the last one that they trained on,

167
00:08:50,200 --> 00:08:52,160
Speaker 3: which again is the world's leading model.

168
00:08:52,320 --> 00:08:53,800
Speaker 4: So we're super excited about that.

169
00:08:54,440 --> 00:08:57,120
Speaker 3: We're landing Trainium to servers now and they're already in

170
00:08:57,160 --> 00:09:00,679
Speaker 3: operation and Nthropic has already is our using parts of

171
00:09:00,679 --> 00:09:01,360
Speaker 3: that cluster, and.

172
00:09:01,320 --> 00:09:02,720
Speaker 4: So super excited about that.

173
00:09:02,760 --> 00:09:04,800
Speaker 3: And the performance that we're seeing out of Trainium too

174
00:09:05,520 --> 00:09:08,760
Speaker 3: continues to be very impressive and really pushes the envelope

175
00:09:08,800 --> 00:09:12,120
Speaker 3: I think on what's possible both from an absolute performance

176
00:09:12,120 --> 00:09:14,880
Speaker 3: basis as well as a cost, performance and scale basis.

177
00:09:15,040 --> 00:09:16,880
Speaker 3: I think some of those are equally going to be

178
00:09:16,960 --> 00:09:19,439
Speaker 3: really important as we move forward in this world, because

179
00:09:19,480 --> 00:09:22,240
Speaker 3: today much of the feedback you get is that AI

180
00:09:22,360 --> 00:09:25,679
Speaker 3: is still too expensive. But costs are coming down pretty aggressively,

181
00:09:25,880 --> 00:09:28,280
Speaker 3: and it's still too expensive, and so we think there's

182
00:09:28,280 --> 00:09:30,680
Speaker 3: a number of things that need to happen there. Innovation

183
00:09:30,760 --> 00:09:32,720
Speaker 3: on the silicon level is one of those things that

184
00:09:32,800 --> 00:09:35,240
Speaker 3: needs to help bring the cost down, as well as

185
00:09:35,240 --> 00:09:38,079
Speaker 3: innovation on the software side and algorithmic side so that

186
00:09:38,120 --> 00:09:41,120
Speaker 3: you have to use less compute per unit of inference

187
00:09:41,440 --> 00:09:44,000
Speaker 3: or training. So all of those are important to bring

188
00:09:44,000 --> 00:09:46,160
Speaker 3: that cost down to make it more and more possible

189
00:09:46,840 --> 00:09:49,120
Speaker 3: for ADI to be used in all of the places

190
00:09:49,320 --> 00:09:52,360
Speaker 3: that we think that it will be over time Matt.

191
00:09:52,440 --> 00:09:56,400
Speaker 2: On Wednesday, Nvidia CEO Jensen Wang summarized inference demand for me.

192
00:09:56,440 --> 00:09:58,160
Speaker 2: I just wanted to play you that SoundBite.

193
00:09:58,320 --> 00:09:59,319
Speaker 4: Sure, well, we.

194
00:09:59,320 --> 00:10:02,440
Speaker 5: Got a whole bunch of engines firing right now. The

195
00:10:02,480 --> 00:10:06,719
Speaker 5: biggest one, of course, is the reasoning AI inference. The

196
00:10:06,760 --> 00:10:12,200
Speaker 5: demand is just off the charts. You see the popularity

197
00:10:12,280 --> 00:10:13,600
Speaker 5: of all these AI services.

198
00:10:13,640 --> 00:10:17,520
Speaker 2: Now your pitch for trainium too. And as you know,

199
00:10:17,559 --> 00:10:19,680
Speaker 2: I've kind of taken a part the serve a design

200
00:10:19,720 --> 00:10:23,360
Speaker 2: and looked at it is the efficiency and cost efficiency

201
00:10:23,440 --> 00:10:27,160
Speaker 2: relative to Nvidia tech. Are you seeing that same demand

202
00:10:27,240 --> 00:10:31,840
Speaker 2: Jensen outlined for Trainium two outside of the relationship with Amthropic.

203
00:10:33,160 --> 00:10:35,680
Speaker 3: Yeah, Look, we're seeing it across a number of different places,

204
00:10:35,679 --> 00:10:38,680
Speaker 3: but it's not really Trainingum two versus in Nvidia, and

205
00:10:38,720 --> 00:10:40,240
Speaker 3: I think that's not really the right way to think

206
00:10:40,240 --> 00:10:42,640
Speaker 3: about it. I think there's plenty of room. The opportunity

207
00:10:42,679 --> 00:10:45,280
Speaker 3: in this space is massive. It's not one versus the other.

208
00:10:45,320 --> 00:10:47,439
Speaker 3: We think that there's plenty of room for both these

209
00:10:47,480 --> 00:10:49,760
Speaker 3: and Jensen and I speak about this all the time

210
00:10:49,960 --> 00:10:53,520
Speaker 3: that in Vidia is an incredibly fantastic platform. They've built

211
00:10:53,520 --> 00:10:57,160
Speaker 3: a really strong platform that's useful and is the leading

212
00:10:57,200 --> 00:11:00,320
Speaker 3: platform for many many applications out there, and so we

213
00:11:00,480 --> 00:11:03,640
Speaker 3: are incredible design partners with them. We make sure that

214
00:11:03,679 --> 00:11:06,240
Speaker 3: we have the latest in video technology for everyone, and

215
00:11:06,320 --> 00:11:08,920
Speaker 3: we continue to push the envelope on what's possible with

216
00:11:09,040 --> 00:11:11,880
Speaker 3: all of the latest in Vidia capabilities. And we think

217
00:11:11,880 --> 00:11:14,880
Speaker 3: there's room for Trainium and other technologies as well, and

218
00:11:14,880 --> 00:11:18,200
Speaker 3: we're really excited about that, and so we have many

219
00:11:18,280 --> 00:11:21,720
Speaker 3: of the leading AI labs are incredibly excited about using

220
00:11:22,000 --> 00:11:24,440
Speaker 3: Trainium too, and really leaning into the benefits that you

221
00:11:24,480 --> 00:11:27,360
Speaker 3: get there, But for the law for a long time,

222
00:11:27,400 --> 00:11:29,920
Speaker 3: these things are going to be living in concert together,

223
00:11:30,120 --> 00:11:32,679
Speaker 3: and I think there's plenty of room, and customers want choice.

224
00:11:32,720 --> 00:11:34,599
Speaker 3: At the end of the day, Customers don't want to

225
00:11:34,640 --> 00:11:37,040
Speaker 3: be forced into using one platform or the other. They'd

226
00:11:37,040 --> 00:11:39,000
Speaker 3: love to have choice in Our job at AWS is

227
00:11:39,040 --> 00:11:41,160
Speaker 3: to give customers as much choice as possible.

228
00:11:42,120 --> 00:11:47,040
Speaker 2: What is general availability of Nvidia GB two hundred for AWS?

229
00:11:47,040 --> 00:11:52,040
Speaker 2: And have you, I guess, launched Grace Blackwell backed instances yet?

230
00:11:52,280 --> 00:11:56,040
Speaker 3: Yes, yep, so we've launched our they would call them

231
00:11:56,040 --> 00:11:59,240
Speaker 3: P six instances, And so those are available in AWS

232
00:11:59,280 --> 00:12:02,000
Speaker 3: today and customers are using them and liking them and

233
00:12:02,040 --> 00:12:04,800
Speaker 3: the performance is fantastic. So those are available today. We're

234
00:12:04,800 --> 00:12:08,640
Speaker 3: continuing to ramp capacity. We work very closely with the

235
00:12:08,720 --> 00:12:12,320
Speaker 3: Nvidia team to aggressively ramp capacity and demand as strong

236
00:12:12,600 --> 00:12:15,720
Speaker 3: for those P six instances. But customers are able to

237
00:12:15,720 --> 00:12:19,440
Speaker 3: go and test those out today, and like I said,

238
00:12:19,720 --> 00:12:22,280
Speaker 3: we're ramping capacity incredibly fast all around the world and

239
00:12:22,400 --> 00:12:24,440
Speaker 3: in our various different regions.

240
00:12:25,800 --> 00:12:30,400
Speaker 2: Now, what is your attitude to Claude Anthropics model being

241
00:12:30,440 --> 00:12:33,480
Speaker 2: available elsewhere on Azure Foundry for example?

242
00:12:35,000 --> 00:12:36,880
Speaker 4: Great I mean that's okay too.

243
00:12:36,920 --> 00:12:40,720
Speaker 3: I think many of our customers make their applications available

244
00:12:40,840 --> 00:12:44,800
Speaker 3: in different places, and we understand that various different customers

245
00:12:44,840 --> 00:12:48,480
Speaker 3: want to use capabilities in different areas and different clouds.

246
00:12:48,920 --> 00:12:51,280
Speaker 3: Our job is to make AWS and this is what

247
00:12:51,320 --> 00:12:54,640
Speaker 3: we do, is to make AWS the best place to

248
00:12:54,760 --> 00:12:58,640
Speaker 3: run every type of workload, and that includes anthropic claud models, but.

249
00:12:58,600 --> 00:13:00,000
Speaker 4: It includes a wide range of things.

250
00:13:00,040 --> 00:13:04,240
Speaker 3: And frankly, that's why we see big customers migrating over

251
00:13:04,280 --> 00:13:08,120
Speaker 3: to AWS. Take somebody like a Mondali's who's really gone

252
00:13:08,160 --> 00:13:11,280
Speaker 3: all in with AWS and moved some of their workloads

253
00:13:11,280 --> 00:13:13,360
Speaker 3: to there. One of the reasons is that they see

254
00:13:13,400 --> 00:13:16,360
Speaker 3: that we have capabilities sometimes using AI by the way,

255
00:13:16,600 --> 00:13:20,120
Speaker 3: in order to really help them optimize their costs and

256
00:13:20,600 --> 00:13:24,160
Speaker 3: have the most available, most secure platform in monthlies. This case,

257
00:13:24,400 --> 00:13:28,120
Speaker 3: they're taking many of their legacy Windows platforms and transforming

258
00:13:28,200 --> 00:13:30,800
Speaker 3: them into Linux applications and saving.

259
00:13:30,480 --> 00:13:32,240
Speaker 4: All of that licensing costs.

260
00:13:32,440 --> 00:13:35,000
Speaker 3: But we have many customers who are doing that, and

261
00:13:35,440 --> 00:13:38,280
Speaker 3: so our job is to make AWS by far the

262
00:13:38,280 --> 00:13:42,400
Speaker 3: most technically capable platform that has the most and widest

263
00:13:42,440 --> 00:13:44,679
Speaker 3: set of services, and that's.

264
00:13:44,480 --> 00:13:44,920
Speaker 4: What we do.

265
00:13:45,400 --> 00:13:47,920
Speaker 3: But I'm perfectly happy for other people to use, Like,

266
00:13:48,120 --> 00:13:51,679
Speaker 3: it's great that Claud's making their services available elsewhere and

267
00:13:52,280 --> 00:13:54,680
Speaker 3: we see the vast majority of that usage happening in AWS.

268
00:13:54,679 --> 00:13:54,880
Speaker 4: Though.

269
00:13:55,679 --> 00:13:58,680
Speaker 2: Will we see open AI models on AWS this year?

270
00:13:59,400 --> 00:14:02,360
Speaker 3: Well, just like you know, we encourage all of our

271
00:14:02,400 --> 00:14:05,400
Speaker 3: partners to be able to be available elsewhere. I'd love

272
00:14:05,440 --> 00:14:06,960
Speaker 3: for others to take that same tack.

273
00:14:08,760 --> 00:14:11,959
Speaker 2: Let's end it with this a question from the audience actually,

274
00:14:12,000 --> 00:14:14,199
Speaker 2: which is where you're going to grow data center capacity

275
00:14:14,280 --> 00:14:16,319
Speaker 2: around the world. I got a lot of questions from

276
00:14:16,440 --> 00:14:19,960
Speaker 2: Latin America and Europe in particular where Jensen flies to

277
00:14:20,080 --> 00:14:20,560
Speaker 2: next week?

278
00:14:20,640 --> 00:14:22,120
Speaker 4: Yeah. Great.

279
00:14:22,880 --> 00:14:26,040
Speaker 3: So in Latin America we're continuing to span expand our

280
00:14:26,040 --> 00:14:29,840
Speaker 3: capacity pretty aggressively. Actually, earlier this year we launched our

281
00:14:29,880 --> 00:14:32,880
Speaker 3: Mexico region, which has been really well received by customers,

282
00:14:33,080 --> 00:14:35,680
Speaker 3: and we've announced a new region in Chile. And we

283
00:14:35,720 --> 00:14:38,080
Speaker 3: already have and for many years have had a region

284
00:14:38,080 --> 00:14:41,160
Speaker 3: in Brazil which is quite popular and has many of

285
00:14:41,360 --> 00:14:44,840
Speaker 3: the largest financial institutions in South America running there. So

286
00:14:45,960 --> 00:14:49,560
Speaker 3: across Central and South America, we are continuing to rapidly expand.

287
00:14:50,200 --> 00:14:52,280
Speaker 3: In Europe we're expanding as well. We have many regions

288
00:14:52,320 --> 00:14:54,520
Speaker 3: already in Europe. One of the things I'm most excited

289
00:14:54,520 --> 00:14:56,680
Speaker 3: about actually is at the end of this year we're

290
00:14:56,680 --> 00:14:59,280
Speaker 3: going to be launching the European Sovereign Cloud, which is

291
00:14:59,320 --> 00:15:02,920
Speaker 3: a unique capability that no one has, which is completely

292
00:15:02,960 --> 00:15:07,840
Speaker 3: designed for critical EU focused sovereign workloads, and we think

293
00:15:08,440 --> 00:15:11,680
Speaker 3: given some of the concerns that folks have around data sovereignty,

294
00:15:11,920 --> 00:15:16,040
Speaker 3: particularly for government workloads as well as regulated workloads, we

295
00:15:16,040 --> 00:15:19,360
Speaker 3: think that's going to be an incredibly op popular opportunity

296
00:15:19,360 --> 00:15:19,880
Speaker 3: for everybody.

297
00:15:20,920 --> 00:15:24,120
Speaker 2: Matt Garman AWSCO, thank you very much.

298
00:15:24,280 --> 00:15:25,120
Speaker 4: Thank you for having me